history of prosthetics and the neurologically controlled arm

advertisement
B1 #6254
Disclaimer — This paper partially fulfills a writing requirement for first year (freshman) engineering
students at the University of Pittsburgh Swanson School of Engineering. This paper is a student, not a
professional, paper. This paper is based on publicly available information and may not be provide
complete analyses of all relevant data. If this paper is used for any purpose other than these authors’ partial
fulfillment of a writing requirement for first year (freshman) engineering students at the University of
Pittsburgh Swanson School of Engineering, the user does so at his or her own risk.
THE SIGNIFICANCE OF THE NEUROLOGICALLY
CONTROLLED ARM
Tim Bowers (teb50@pitt.edu) Jeremiah Hoydich (jsh70@pitt.edu)
Abstract- Prosthetics have been around for centuries.
Whenever we think of prosthetics we typically think about
plastic molds resembling limbs that do not have many
functions. Scientists now are working to change that.
Prosthetics are now being designed and tested that use the
power of the brain to bring the functions of the arm back to
amputees and quadriplegics [1]. These neurologically
controlled arm prosthetics will be a critical part of the future
of prosthetics.
Neurological prosthetics use electrodes along the natural
pathways of the brain to control simple motions of a
mechanical arm [2]. As the technology improves, so does the
intricacy of the movements the arms are able to make.
Currently, the only drawbacks with this technology are the
relatively high cost and inability for patients to get the
technology through health insurance and Medicare [3].
Key Words- Electrode Technology, Motor Cortex,
Neuroprosthetic, Posterior Parietal Cortex, Prosthetics,
Quadriplegic.
THE IMPORTANCE OF THE
NEUROPROSTHETIC ARM
Each year in the United States an estimated 185,000 people
have one limb amputated [4]. The number of amputees in the
United States is projected to be 3.6 million by 2020 [4].The
number of people who do not have function of their limbs only
increases when you take into account the number of people
with tetraplegia and other forms of degenerative spinal cord
diseases. For these people, normal everyday life will likely
never be a possibility again. This is especially true for people
who have had arm amputations or loss of arm movement. The
prosthetics provided today for arm amputees can only do the
basic functions of grasping and arm movement. And people
with degenerative diseases cannot even use these rudimentary
prosthetics. Both of these people could be helped with the
advent of the neurologically controlled prosthetic arm. This
prosthetic would allow the user to control a fully functioning
mechanical arm and hand the same way they controlled their
old limbs: with their brain. They large scale introduction of
these prosthetics could bring back the old everyday lives to
these patients.
HISTORY OF PROSTHETICS
The history of the arm prosthetic starts in the historical
writing of Pliny the Elder [5]. Pliny writes about a Roman
general who had his arm amputated during the Second Punic
War [5]. To keep fighting, the general had a metal arm cast
with which he could hold his shield [5]. The next advancement
happened during the Dark Ages when the hook was added to
the prosthetic arm to give it some functionality in being able to
pick up objects [5]. The first functioning hand prosthetic was
crafted 1508 for a German mercenary [5]. This prosthetic
could be made to grasp objects by manipulating it with your
freehand [5]. Since then prosthetics have been better molded
to the patient’s arm and have a better hook mechanism but
these changes have not had a huge change in the area of arm
prosthetics. The neuroprosthetic arm hopes to change this.
The neuroprosthetic arm has been a long time imagined
advancement in prosthetics. As most modern technology, the
premonitions of this prosthetic started in science fiction. This
prosthetic was popularized by Star Wars. After Luke
Skywalker gets his hand cut off by his father Darth Vader, he
gets a fully mechanical prosthetic arm and hand which are
controlled with his mind. The actual science behind the
neuroprosthetic arm was not figured out until 2002 when Dr.
Kevin Warwick had electrodes implanted in his brain which
allowed a neuroprosthetic arm to mimic the movements his
natural arm was making [6]. From this point many companies
and colleges, including the University of Pittsburgh, have
joined in the process of creating the ideal neuroprosthetic arm.
There have been many advancements. Here are two of the
more intriguing advancements researchers have made for this
type of prosthetic. First, a patient at Case Western University
was successfully able to use two neuroprosthetic arms at the
same time [4]. This is a huge advancement, as these people
would be as close as possible to life before their amputation or
illness. The other notable advance also takes place at Case
Western. In this study, a man who lost a hand was able to feel
when a cotton-ball was dragged across a neuroprosthetic hand.
This advance is incredibly interesting as not only can the
University of Pittsburgh, Swanson School of Engineering 2016-02-12
Timothy Bowers, Jeremiah Hoydich
device decipher the electrical activity in a patient’s brain, but
it can also relay information to the brain to trigger a biological
response.
These advancements were over a fourteen year time span. For
science, this is a relatively short time which makes you wonder
what heights this technology can reach if the advancements
keep up at this speed.
representation than the trunk or legs because the muscle
patterns are relatively simple [7].
Secondary Motor Cortices
The secondary motor cortices are involved in motor
planning. These cortices are consisted of three parts: the
posterior parietal cortex, the premotor cortex, and the
supplementary motor area (SMA). The posterior parietal
cortex is involved in transforming visual information into
motor commands. It sends information to the premotor cortex
and supplementary motor area. For example, the posterior
parietal cortex is involved in determining how to steer the arm
to a glass of water based on where the glass of water is located
in space.
The premotor cortex lies just in front of or anterior to the
primary cortex. It helps to guide body movements by
integrating sensory information. It controls the muscles that
are closest to the body’s main axis. For example, the premotor
cortex would help orient the body before reaching for the glass
of water [7].
The supplementary motor area (SMA) lies above or medial
to the premotor cortex and in front of or anterior to the primary
cortex. The SMA is involved in planning complex movements
and in coordinating movements involving both hands. Both the
SMA and premotor cortex send information to the primary
motor cortex as well as to brainstem motor regions [7].
THE ANATOMY OF MOVEMENT
Cerebellum
The cerebellum is a small grooved structure located in the
back of the brain beneath the occipital lobe. It is involved in
the timing and coordination of motor programs. Motor
programs are generated in the basal ganglia. The basal ganglia
is several subcortical regions that are involved in organizing
motor programs for complex movements, such as learning to
play a new sport or instrument. The basal output sends an
output to other subcortical brain regions and to the motor
cortex [7].
MOTOR CORTEX
The motor cortex controls our body’s voluntary
movements. It must receive various kinds of information from
the various lobes of the brain. From the parietal lobe, it
receives information about our body’s spatial awareness. The
information on the goal at hand and appropriate strategies for
obtaining it come from the anterior frontal lobe. The temporal
lobe sends memories of past strategies to the motor cortex to
perform the task at hand.
The motor cortex then generates neural impulses that
control the execution of movement. These signals cross the
midline to activate skeletal muscles on the opposite side of the
body. This process allows the right hemisphere to control the
left side of the body and the left hemisphere to control the right
side [8].
The motor cortex lies along the precentral gyrus in the rear
portion of the frontal lobe, just before the central sulcus, or
furrow. The central sulcus (furrow) separates the frontal lobe
from the parietal lobe. The motor cortex is divided into two
areas: the primary motor cortex and secondary motor cortices
[8].
Corticospinal Tract
The corticospinal tract is the main pathway for control
of voluntary movement in humans. This is the only direct
pathway from the cortex to the spine. Neurons in the brain give
rise to over a million fibers in this tract. These fibers descend
through the brainstem and cross over to the opposite side of the
body. The fibers continue to descend through the spine until
ultimately terminating at the appropriate spinal levels [7].
Cortical Control of Skeletal Muscles
Signals generated in the primary cortex travel down the
corticospinal tract, through the spinal white matter to synapse
on interneurons and motor neurons in the spinal cord’s ventral
horn. The ventral horn sends its axons out through the ventral
routes to innervate individual muscle fibers.
The ventral horn neuron, its axon, and myofibrils that it
innervates are considered to be all a single motor unit. Motor
neurons can innervate any number of muscle fibers, but each
fiber is only innervated by one motor neuron. When the motor
neuron fires, all of its innervated muscle fibers contract. The
size of the motor units and the number of fibers that are
innervated contribute to the force of the muscle contraction [7].
There are two types of motor neurons in the spine: alpha
and gamma motor neurons. Alpha motor neurons innervate
muscle fibers that contribute to force production of the muscle.
Primary Motor Cortex
The primary motor cortex forms a thin band along the
central sulcus (furrow). Every part of the body is represented
in the primary cortex in what is known as the motor
homunculus. The amount of brain matter devoted to any
particular body part represents the amount of control that the
primary motor cortex has over that body part. For example, a
lot of cortical space is required to control complex movements
of the hands and fingers. These body parts have a larger
2
Timothy Bowers, Jeremiah Hoydich
Gamma motor neurons measure the length, or stretch, of the
muscle [7].
This complex process allows us humans to perform tasks
effortlessly and learn new tasks everyday.
Neurofeedback is the decoded motor intentions of the
patient. These intentions are used in real-time to control an
assistive device. A decoding algorithm is calibrated using
neural signals collected during performance of an instructed
set of real or imagined arm movements. The parameters of
movement are tuned such that the output of the algorithm best
predicts the direction of the training movements [2].
When the output of the algorithm during this phase does
not influence the user’s behavior, it is known as one-loop
decoding [2]. The decoding algorithm may be used for realtime brain control, during which the user receives some form
of feedback from the device. This is known as closed-loop
decoding.
Proprioceptive information, information relating to stimuli
that are produced and perceived within an organism, can
improve the accuracy of the BMI-meditated movement. Once
the brain is “in the loop” and connected to the decoding
algorithm, patterns of neural activity change as the user
masters the interface.
Individuals learn voluntary control of a feedback signal that
relays real-time information about specific brain activity.
Practice using the BMI is accompanied by profound
reorganization of cortical activity, such as changes in the
directional tuning of neurons used by the decoding algorithm
and reduction of the modulation of neighboring neurons [2].
Improvements to the neurofeedback could involve residual
sensory models, such as vision or artificial sensory pathways
provided by electrical stimulation to the brain. By
incorporating the senses into the interface will allow the user
to have not only the functionality to think to perform a task,
but also the awareness to sense their surrounds to perform it
more easily. Repetitive stimulation might induce long-term
changes that increase the excitability of spinal circuitry and
enhance the efficacy with which movements can be evoked
[2].
INTERFACE WITH THE BRAIN
Research has seen rapid progress in two examples of
neuroprosthetics for spinal cord injuries (SPI). These examples
are brain-machine interfaces (BMI) and functional electrical
stimulation (FES). A BMI enables a patient to control assistive
devices, such as robotic limbs, by using neural signals
recorded directly from the brain. FES is used in the attempt to
reanimate paralyzed limbs. [2]
Brain-Machine Interfaces
Implanted neurostimulators may find applications in
modulating the excitability of spinal networks and guiding the
activity-dependent processes that govern the formation of new
motor circuits. The BMI record and decode signals from the
brain enabling voluntary control of assistive devices. It
modifies patterns of cortical activity through the process of
neurofeedback [2].
Invasive techniques of direct brain stimulation have been
used. BrainGate, a brain control system that uses an array of
96 silicon electrodes that penetrate 1.5mm into the upper-limb
representation of the motor cortex to record firing from 50 or
more neurons. Spiking rates of these neurons are then
processed to provide control signals for various artificial
effectors [2].
Noninvasive recording techniques, such as an
electroencephalogram (EEG), are an alternative method to
obtain signals for neural interfacing. An EEG is a test that
detects electrical activity in your brain using small, flat metal
discs (electrodes) attached to one’s scalp. One’s brain cells
communicate via electrical impulses and are active all the time,
even when one is asleep [2].
Unfortunately, as any imagined movement of a given limb
produces the same general pattern, desynchronization over a
wide area of cortex, interferes with the specificity of what
signals are going toward the movement of which limb. The
independent control of movements in multiple dimensions
requires the patient to learn nonintuitive combinations of lefthand, right-hand, and foot movements. The EEG is generally
poor at capturing the high-frequency bands. This is possibly
due to the spatiotemporal filtering that is inherent in scalp
recordings. Scalp recordings use spatiotemporal filtering to
determine the active areas in the brain when given a task to
perform. Also, low frequency EEG signals may sometimes be
confounded by eye movements that cause occipital lobe
activity. This activity can interfere with the intention of
recording motor signals [2].
MECHANICS OF THE PROSTHESIS
Robotic arm prosthetics come in many different shapes and
sizes. Often a design team will sacrifice design appeal to
achieve greater function. The relative importance of the
appearance as opposed to the functionality is dependent on the
patient. The components of a full arm prosthetic are the wrist,
forearm, upper-arm, elbow, and shoulder (scapular). The most
common actuator for electrically powered prosthesis is the
permanent magnetic dc electric motor with some form of
transmission [9].
The prosthesis can experience motion in either two or
three dimensions dependent upon the parameters set. Two
dimensional motion allows the prosthesis to only move on one
plane, whether it be vertical or horizontal. Three dimensional
motion of a full arm prosthetic includes: wrist flexionextension, wrist rotation, wrist abduction-adduction, upperarm flexion-extension, upper-arm rotation, upper-arm
Neurofeedback
3
Timothy Bowers, Jeremiah Hoydich
abduction-adduction, elbow flexion-extension, shoulder
abduction-adduction, and shoulder elevation-depression [9].
An experiment done by the Department of Physiology and
Pharmacology at State University of New York Downstate
Medical Center in Brooklyn, New York involved the
interconnection of a 3-layered cortex, composed of several
hundred
spiking
model-neurons,
which
display
physiologically realistic dynamics, to a two-joint
musculoskeletal model of a human arm. The model of the
human arm is composed of realistic anatomical and
biomechanical properties. The virtual arm received muscle
excitations from the neuronal model and fed back
proprioceptive information, forming a closed-loop system.
The virtual arm muscle activations responded to motor neuron
firing rates, with virtual arm muscle lengths encoded via
population coding in the proprioceptive population. The
researchers calibrated the robotic arm to reproduce the same
trajectories in real time and compared the dimensionless jerk
measures between the musculoskeletal arm and a simple arm
design [10].
The virtual arm includes rigid bodies (bones), joints,
muscles, and tendons. The kinematics are governed by a set of
ordinary differential equations (ODEs) that compute muscle
activation, length, force, as well as arm motions and forces at
millisecond resolution. The cortical model was interfaced with
the virtual arm by exciting the arm muscles using a spiking
output from motor neuron output. The proprioceptive
information from the muscle lengths provided activation for a
proprioceptive neural population. Arm joint angles were also
fed back to the biomimetic model and was used to calculate the
error signal during the reinforcement learning-based training
phase [10].
The kinematics of each joint and the force-generating
parameters for each muscle in the system in a biomechanical
model of the upper extremity musculoskeletal system have
been derived by anatomical and psychological studies
represented an average size adult human male. The model
captured the primary features of the upper extremity geometry
and mechanics. This includes the complex joint coupling
effects, where the mechanics of a given joint depends on the
posture of the adjacent joints. The model included the
following rigid bodies where the muscles are anchored:
ground, thorax, clavicle, scapula, humerus, ulna, radius, and
hand [10].
This experiment only allowed for two degrees of motion,
which was shoulder and elbow joint rotation in the horizontal
plane. The major active muscles in shoulder and elbow motion
are: the posterior deltoid, infraspinatus, lattisimus dorsi, and
teres minor (shoulder exterior muscles); anterior deltoid,
pectoralis major, and corachobrachialis (shoulder flexor
muscles); triceps (elbow extensor muscles); biceps and
brachialis (elbow flexor muscles). Other muscles were set for
joint stability, these included: lateral deltoid, anconeous,
brachioradialis, extensor carpi radialis longus, and pronator
teres. Muscles with multiple heads had the muscle branches
connected to different insertion and origin points but were
controlled by the same input signal for simplicity.
As a result of the experiment, the mean dimensionless jerk
measure was significantly lower for the realistic arm as
compared to the simple arm, both in trained and naïve
networks. This suggests that the musculoskeletal arm for any
biologically reasonable input generates smoother movements
than the simple arm [10].
This experiment demonstrated that increasing the realism
of the arm model reduces the arm trajectory jerk and results in
velocity profiles closer to biology, which reflect into smoother
robot movements.
CASE STUDIES
There have been a number of real world tests of the
neurologically controlled arm. Here are some of the finding
that these different researchers have found. The first study we
will talk about was conducted by the Applied Physics
Laboratory of John Hopkins University. In this study a 52 year
old women with tetraplegia is tested through various tasks
using an anthropomorphic prosthetic limb [3]. To start the
study, the researchers placed two intracortical microelectrode
arrays into the left motor cortex of the patient [3]. Over the
course of three weeks the patient was trained in the use of the
arm, first in the general movement of the arm and then the use
of the fingers and joints. After this training, the patient came
back and did nine tests out of 19 possible tests a day for 98
days [3]. The tests included activities such as stacking cones
and moving block a certain distance. Each test was timed. The
patient was graded on completion and the time it took to
complete the task. Over the course of the test the results of the
patients steadily improved.
This case study illustrates how these prosthetics can
perform tasks for people who normally would never be able to
do them. This could be a huge step in overcoming their
disabilities.
Another case study has been performed by the University
of Pittsburgh in which monkeys have been trained to control a
virtual arm to grab a point in space on a computer with their
mind [11]. Although this study is not as far along as the
previous study, the promising point here is that even monkeys
have the ability to operate the device.
The University of Pittsburgh expanded on their work with
neuroprosthetics when they tested a neuroprosthetic arm on a
52 year old female patient with tetraplegia [12]. To get the
prosthetic to work, researchers examined the patient’s brain
pattern as they asked her to move her arm and hand in a certain
way [12]. They then programed the prosthetic to make those
motions and handshapes as it sensed the patient making them.
For calibration, the researchers hooked the patient into a
computer and asked her to do the same tasks in virtual reality.
This study went further than most other studies in the number
of ways the hand could be manipulated [12]. Previous studies
only had six hand formations to choose from, this study tried
increasing this number to ten [12]. The patient was
4
Timothy Bowers, Jeremiah Hoydich
successfully able to manipulate all the additional handshapes
[12]. Although these advancements in the number of
handshapes is a good step forward the functionality of the
prosthetic suffered. The ability to hold and move objects was
inconsistent through the first round of tests [12]. As the
researchers improved the process of calibration the success
rate of object manipulation rose [12]. The success rates of this
study were surprisingly good. Success rate of tasks by the
patient were 70% [12]. Overall, this study illustrates the bright
future of the neuroprosthetic arm. This study was able to
mimic more hand motions than before which means that these
prosthetics are on the path to becoming as fully functional as a
regular arm.
These case studies show how neurological devices are not
just a science fiction fantasy anymore. They are science fact.
Fact that deserves further funding and research because of the
help that it can bring to potentially millions of people.
FROM IMAGINATION TO REALITY
Neurologically controlled prosthetics is a stepping- stone
into the future of rehabilitation science. It gives people the
chance to perform simple tasks that once seemed impossible to
them. The history of prosthetics shows the drastic progression
of a field that’s sole purpose is to improve the lives of those in
need. The history of arm prostheses identifies the increasing
development and potential this technology is moving towards
to help people. The prosthetics improvement over time has
garnered its way to an astounding feat.
The brain’s motor cortex and secondary cortices have given
people the ability to do numerous tasks in their everyday lives
from the simple task of picking up a glass of water to the
complex tasks of playing an instrument.
The use of the brain to control a machine has been thought
of and imagined throughout history and the technology to do
so is astounding. The neurologically controlled prosthetic arm
has allowed an entirely new group of people to experience a
form of independence after a part of their independence has
been taken from them.
By tapping into the brain’s motion capabilities of the
human body people who have lost the control of movement in
their arms are now able to provide movement to artificial arms
to perform their everyday tasks. A part of their independence
is finally being able to be given back to them.
The case studies show that this new technology is paving a
way to a new future for those who are trying to regain the
independence in their everyday lives. Along with this new
technology, ethical concerns arise. The positives seem to far
outweigh the negatives as the main purpose is to benefit the
lives of those in need.
The idea of neurologically controlled devices is stepping
out of what was only thought possible in the imagination into
the reality of our everyday lives.
ETHICS
There are currently three ethical dilemmas facing the future
of the neuroprosthetic arm currently. These three dilemmas are
the relatively high cost of the prosthetic, the cybersecurity risk,
and the invasiveness of the procedure.
Currently, Medicare will only cover advanced prosthetics
for leg amputees who they believe have a high level of mobility
[4]. And if a patient would want a neuroprosthetic arm that is
fully functional the cost for the prosthetic alone can be
upwards of $80,000 [4]. To combat the high cost of these
prosthetics an advocacy group named the Amputee Coalition
has been lobbying for Medicare to cover a broader range of
patients.
Cybersecurity has also been a growing risk in the health
field in general. With increasing computer reliance the threat
of hacking medical devices has also risen. Neuroprosthetics
are no exception. Their reliance on computers makes them an
easy target for hackers [13]. These prosthetics likely don’t
have any countermeasures to these hacks as they are not often
expected. This can leave the user in a dangerous situation as
hacking the prosthetic and disabling it during a dangerous task
could be potentially deadly.
The last ethical dilemma is the invasiveness of the
procedure. For the neuroprosthetic to work, electrodes need to
be placed within the patient’s brain. For amputees and patients
suffering from tetraplegia this is not a problem, but for stroke
victims this can be an issue. Since some stroke victims can
regain the use of their arms through physical therapy the
invasiveness of the procedure could potentially harm their
recovery because of the location of the electrodes on the areas
of the brain which control limb movement [4]. There have been
improvements to get rid of the invasiveness of the procedure.
Researchers at the University of Houston have developed a cap
that reads brain activity the same way the invasive procedure
does and negates the need for the electrodes to be planted
within the brain [4].
REFERENCES
[1] T. Yanagisawa. (2011). “Electrocorticographic control of a
prosthetic arm in paralyzed patients.” Annals of Neurology.
(Online
article).
http://onlinelibrary.wiley.com/doi/10.1002/ana.22613/ful.
[2] A. Jackson. (2012). “Neural interfaces for the brain and
spinal cord-restoring motor function.” Nature Reviews
Neurology
8.
(Online
article).
http://www.nature.com/nrneurol/journal/v8/n12/full/nrneurol.
2012.219.html.
[3] J. Collinger (2013). “High-performance neuroprosthetic
control by an individual with tetraplegia.” The Lancet. (Online
article).
http://www.sciencedirect.com/science/article/pii/S014067361
269.
5
Timothy Bowers, Jeremiah Hoydich
[4] J. Jacob. (2015). “Advance Prosthetics Provide More
Functional Limbs”. The Journal of the American Medical
Association. (Online journal).
http://jama.jamanetwork.com/article.aspx?articleID=2319161
[5] K. Norton. (2007). “A Brief History of Prosthetics”.
Amputee Coalition. (Online article). http://www.amputeecoalition.org/resources/a-brief-history-of-prosthetics/.
[6] K. Warwick. (2003). “The Application of Implant
Technology for Cybernetic Systems”. The Journal of the
American Medical Association. (Online journal).
http://archneur.jamanetwork.com/article.aspx?articleid=7847
43.
[7] S. Schwerin. (2013). “The Anatomy of Movement” Brain
Connection.
(Online
article.).
http://brainconnection.brainhq.com/2013/03/05/the-anatomyof-movement/
[8] “The Motor Cortex” THE BRAIN FROM TOP TO
BOTTOM.
(Website.)
http://thebrain.mcgill.ca/flash/i/i_06/i_06_cr/i_06_cr_mou/i_
06_cr_mou.html
[9] R. F. ff. Weir. 2004. Standard Handbook of Biomedical
Engineering and Design. McGraw-Hill; [2004; 2016].
[10] S. Dura-Bernal, X Zhou, S. A. Neymotin, A. Przekwas, J.
T. Francis, W. W. Lytton (2015). “Cortical Spiking Network
Interfaced with Virtual Musculoskeletal Arm and Robotic
Arm” Frontiers in Neurobotics. (Online article.).
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4658435/
[11] M. Velliste. (2014). “Motor Cortical Correlates of Arm
Resting in the Context of a Reaching Task and Implications
for Prosthetic Control”. The Journal of Neuroscience. (Online
journal).
http://motorlab.neurobio.pitt.edu/pub/MotorCorticalCorrelate
sofArmResting.pdf.
[12] B. Wodlinger. (2015). “Ten-dimensional
anthropomorphic arm control in a human brain-machine
interface: difficulties, solutions, and limitations”. Journal of
Neural Engineering. (Online journal).
http://iopscience.iop.org/article/10.1088/17412560/12/1/016011/meta;jsessionid=C5E7D3D74BF5B3C923
45D3F4A3012A6E.c4.iopscience.cld.iop.org.
[13] J. Hsu. (2014). “Feds Probe Cybersecurity Dangers in
Medical Devices”. IEEE Spectrum. (Online article).
http://spectrum.ieee.org/tech-talk/biomedical/devices/fedsprobe-cybersecurity-dangers-in-medical-devices.
ACKNOWLEDGEMENTS
We would like to thank our co-chair, Ms. Jesse Liu, for her
help in guiding us in this assignment. Our writing instructor,
Professor Janet Zellman, for providing us with critique on our
writing. And the University of Pittsburgh librarians who
provide us with many avenues to do research.
6
Download